The difficult discussion chapter



The discussion chapter is the problem child of the thesis. The chapter most likely to provoke fear, uncertainty and doubt. Not everyone writes a chapter called “discussion”, but everyone has to do discussiony bits because, well – that’s where the creative magic of the PhD happens.

The discussion section is scary because you have to make new knowledge claims of your own, not just agree or disagree with other people. Knowledge claims are like dumplings in the thesis soup or chocolate chips in the PhD cookie. Without knowledge claims you don’t have a thesis at all, just a report of work that was done.

It’s important to get the discussion bits right. According to the seminal paper by Mullins and Kiley “It’s a PhD, not a Nobel Prize“, examiners want to know if you would be an interesting future colleague. They will evaluate the quality, amount and – most importantly – the believability of your knowledge claims. If you are not sufficiently speculative there will  not be not enough ‘philosophy’ in your thesis soup to justify the title of Dr. If you are too speculative you might tip over the believability cliff.

There’s a lot of advice about why a discussion chapter is important and what it should contain, but relatively little on the mechanics of actually doing it. Years ago I wrote a post on how to start the discussion chapter. I gave a few suggestions for generating creative ideas, but not a detailed discussion of how to do any of them. The post was a useful starting point, but unfortunately not enough to actually help one of my students, Wendy, write her discussion chapters.

Like many thesis writers, Wendy has had a long slog with this project and is having trouble seeing the wood for the trees. Wendy is in the creative arts, so her findings are not in the form of formulas and graphs. The findings are really a series of observations which can be backed up with evidence.

We decided to use the ‘big list’ method described by Evans, Gruba and Zobel in “How to write a Better Thesis” but with extra Inger- style diagrams. I think we made good progress with a whiteboard in the last meeting and I asked her if I could share what we did with the world, she generously agreed (thanks Wendy!). Here’s what we did:

Step one: the mud map

How to write a Better Thesis suggests that you start the discussion by simply writing a huge list of everything you learned about your topic – these are your provisional conclusions, observations, interesting facts and statements. The book suggests that you write this list with the help of a sympathetic listener who knows the project, like a supervisor.

Wendy sat in a chair and just talked at me. I wrote what she said on the whiteboard, asking a question now and then to clarify or expand on a point. The whiteboard was an ideal medium for this task. We could both see it and it was easy to rub out stuff. At the end of about 20 minutes we had this ‘mud map’ of possible findings:


The mud map enabled us to start sorting and organising the ideas. I put a little ‘O’ next to every finding I thought represented an original knowledge claim. Wendy still has one data collection round to run so some findings are more speculative than others. I put ticks next to statements I felt she had sufficient data to argue, question marks next to statements where we are waiting for data and 1/2 wher we weren’t sure. I added a few arrows and brackets to connect obviously related statements.

Step Three: Find themes

The next step is a similar method to so called ‘grounded theory’ coding which is a common method in social science. This is a creative process which involves reading over the mud map again and again, writing ideas for over arching themes under which the findings might fit. After about another 45 minutes of fiddling and arguing about it, we had this provisional list of four themes:

diccicult2We took a couple of photos of the board with my phone and then rubbed it all out.

Step four: use a spider diagram to make connections explicit

This is where we diverted from How to write a better thesis and brought in my favourite thinking tool, the spider diagram. We put the overall aim of the thesis in the middle of the bubble and put the emergent themes in the first layer of bubbles around it:


The third layer of bubbles contained findings from our mud map (recorded in the photos on my phone) that we thought fell under these themes. As we wrote the findings into the bubbles we shifted the wording to find shorter ways of saying the same thing and started to blend some of the findings together.

This process enabled us to see how some findings from the mud map might fall under multiple themes, which suggested a satisfactory repetition was emerging. You know you have a thesis on your hands when you feel like you are saying the same thing over and over, but in a slightly different way every time.

A spider diagram is a great tool for this kind of work because it doesn’t commit you to a structure. Wendy might decide that ‘big data’ should be its own theme, or she might decide to sprinkle the ‘big data’ goodness throughout the other themes. This is a stylistic choice – there is no right or wrong way to do it, but the diagram lets you imagine different results and discuss the implications of the various choices. We took several photos as this diagram evolved and changed to present different options.

Step Five: Make a snowflake

At this point I was happy. I know I can work from a spider diagram, but Wendy was still not comfortable. She is a more creative person than me and was worried that she would muck around and just find more and more connections – a fair point. So I suggested we try using a matrix. While a spider diagram encourages you to see connections, a matrix forces you to think about hierarchy.

The snowflake method is a technique used by novel writers. It involves making a simple outline, which becomes more and more complex as the writing develops. We started by drawing a table like this:

difficult4The first column contains themes from our bubble diagram, the next number of words for that section (bearing in mind that we don’t want this chapter to be more than about 8000 words long). The last column is a list of subheads.

In the row ‘shareability’ there are four provisional subheadings. Each of these subheads will have 500 words – a page of text. To further develop the snowflake, Wendy will write at least four points under each subhead, to make paragraph headings. With a little manipulation these paragraph headings become topic sentences for each paragraph. From a huge mess of ideas the writing task now looks much simpler – just a matter of picking off the paragraph you want to tackle and writing 200 – 300 words.

Of course, in practice writing is not that easy, but this process does help you see the wood instead of the trees. Wendy went away smiling from our meeting and that’s what matters to me most anyway.